In these slides, we will walk through an innovative new approach to machine learning that seeks to model and learn customer choice patterns and preferences from sparse transactional data.
Trailing 6-weeks allows us to normalize any macro-economic factors (like a big store closed across the street, causing sales to go up)
We were working with about 30% of the store spend.
We actually predicted extremely close to the actual. This was not buy cannibalizing sales elsewhere, but with positive spillover effects of a better assorted store. So, the entire store yielded 3% increase.